2.4.21-pre2 IDE problems

2.4.21-pre2 IDE problems

[Marijn Ros]

Hi all,

My (old, but still newest) machine has a UM8886A south bridge with IDE
controller. This chip is not supported and therefor runs at PIO mode
0, yielding a whopping 2MB/s throughput with the normal (old) 2.4 IDE
code.

However, when I try the new 2.4.21-pre2 taskfile IO setting, I get a
'hda: lost interrupt' message every 30 seconds (the timeout period I
guess) during disk IO, making the machine unusable. I know this
setting is experimental, but I guess you would like to know about my
problems before the old code is phased out completely.

I understood that taskfile IO is a backport of 2.5 code, so I decided
to try 2.5.53 too. Strangely, it doesn't loose IDE interrupts, but
seems to loose clock interrupts during disk IO, as ntpd looses sync
(0.6-1.2s per 20 minutes) during a kernel compile but not during
normal usage (web-browsing and e-mail on an otherwise idle
system). Setting kernel preemtion doesn't make a difference (don't
know why I decided to test that anymore).

I am not on this list, but plan to read the archives avery day. If
more information is needed, please CC me so I can send a proper reply.

Bye,
        Marijn

Re: 2.4.21-pre2 IDE problems

[Alan Cox]

On Tue, 2002-12-31 at 12:18, Marijn Ros wrote:
> However, when I try the new 2.4.21-pre2 taskfile IO setting, I get a
> 'hda: lost interrupt' message every 30 seconds (the timeout period I
> guess) during disk IO, making the machine unusable. I know this
> setting is experimental, but I guess you would like to know about my
> problems before the old code is phased out completely.

Taskfile I/O as opposed to ioctl is broken for PIO. I know about this
and I've disabled it. I don't plan to fix that path for 2.4 but to keep
the old read/write paths to reduce risk